Both Firefox and WebKit-based browsers fail a couple of tests because of the way the flex-grow and flex-shrink properties are serialized.

For example, the specified value is using 2.34e+06 notation while the computed value is 2340000. I think the specs should allow both representations as valid, mainly because of the way engines work. If we only allow the specified value as valid representation then we'll be forcing engines to store that value (or something that refers to the actual notation used to represent the number) just for the getComputedStyle() call.

I have a PR open since April for WPT tests but I believe the spec editors are the ones to make the call and decide whether or not to allow those multiple representations.